Learning Pose Invariant and Covariant Classifiers from Image Sequences

Object tracking and detection over a wide range of viewpoints is a long-standing problem in Computer Vision. Despite significant advance in wide-baseline sparse interest point matching and development of robust dense feature models, it remains a largely open problem. Moreover, abundance of low cost mobile platforms and novel application areas, such as real-time Augmented Reality, constantly push the performance limits of existing methods. There is a need to modify and adapt these to meet more stringent speed and capacity requirements. In this thesis, we aim to overcome the difficulties due to the multi-view nature of the object detection task. We significantly improve upon existing statistical keypoint matching algorithms to perform fast and robust recognition of image patches independently of object pose. We demonstrate this on various 2D and 3D datasets. The statistical keypoint matching approaches require massive amounts of training data covering a wide range of viewpoints. We have developed a weakly supervised algorithm to greatly simplify their training for 3D objects. We also integrate this algorithm in a 3D tracking-by-detection system to perform real-time Augmented Reality. Finally, we extend the use of a large training set with smooth viewpoint variation to category-level object detection. We introduce a new dataset with continuous pose annotations which we use to train pose estimators for objects of a single category. By using these estimators' output to select pose specific classifiers, our framework can simultaneously localize objects in an image and recover their pose. These decoupled pose estimation and classification steps yield improved detection rates. Overall, we rely on image and video sequences to train classifiers that can either operate independently of the object pose or recover the pose parameters explicitly. We show that in both cases our approaches mitigate the effects of viewpoint changes and improve the recognition performance.

Object Classification and Detection in High Dimensional Feature Space

Charles Dubout

Object classification and detection aim at recognizing and localizing objects in real-world images. They are fundamental computer vision problems and a prerequisite for full scene understanding. Their difficulty lies in the large number of possible object positions and the appearance variations of object classes. This thesis improves upon several classical machine learning algorithms, enabling large computational gains in high dimensional feature space. A common trend in machine learning and computer vision research is to go large scale. In particular, the advent of huge datasets mined from the Internet, and the combination of multiple feature sources have considerably broadened the applications of computer vision. Tasks which were thought impossible a few years ago, such as human action recognition or pose estimation, automatic outdoor navigation, etc., now seem within reach. This dissertation is divided into two parts. The first one deals with the efficient training of a classifier or detector based on a large number of feature extractors, outside the control of the learning algorithm, and therefore of unknown suitability to the task at hand. More precisely, this part presents two kinds of strategies to accelerate the training of Boosting algorithms in such a context: (a) a method to better deal with the increasingly common case where features come from multiple sources (e.g. color, shape, texture, etc., in the case of images) and therefore can be partitioned into meaningful subsets; (b) new algorithms which balance at every Boosting iteration the number of weak learners and the number of training examples to look at in order to maximize the expected loss reduction. Experiments in image classification and object recognition on four standard computer vision datasets show that the adaptive techniques we propose outperform both basic sampling and state-of-the-art bandit methods. The second part deals with linear object detectors, currently the most popular class of detection systems, encompassing template matching, deformable part models, poselets, convolutional neural networks (which internally use linear filters), etc. The main bottleneck of many of those systems is the computational cost of the convolutions between the multiple rescalings of the image to process and the linear filters. We make use of properties of the Fourier transform and clever implementation strategies to obtain a speedup factor proportional to the filter size, both while training and at test time. We also introduce a few modifications to the original Deformable Part Model (DPM) of Felzenszwalb et al. improving its detection accuracy. The gains in performance are demonstrated on the well-known Pascal VOC benchmark, where an increase by one order of magnitude in the speed of said convolutions, and an average improvement of 15% in the accuracy of the detector are established.

Programme doctoral en Informatique, Communications et Information2013

User Authentication via Adapted Statistical Models of Face Images

Samy Bengio

It has been previously demonstrated that systems based on local features and relatively complex statistical models, namely 1D Hidden Markov Models (HMMs) and pseudo-2D HMMs, are suitable for face recognition. Recently, a simpler statistical model, namely the Gaussian Mixture Model (GMM), was also shown to perform well. In much of the literature devoted to these models, the experiments were performed with controlled images (manual face localization, controlled lighting, background, pose, etc.); however, a practical recognition system has to be robust to more challenging conditions. In this article we evaluate, on the relatively difficult BANCA database, the performance, robustness and complexity of GMM and HMM based approaches, using both manual and automatic face localization. We extend the GMM approach through the use of local features with embedded positional information, increasing performance without sacrificing its low complexity. Furthermore, we show that the traditionally used Maximum Likelihood (ML) training approach has problems estimating robust model parameters when there is only a few training images available; considerably more precise models can be obtained through the use of Maximum a Posteriori (MAP) training. We also show that face recognition techniques which obtain good performance on manually located faces do not necessarily obtain good performance on automatically located faces, indicating that recognition techniques must be designed from the ground up to handle imperfect localization. Finally, we show that while the pseudo 2D HMM approach has the best overall performance, authentication time on current hardware makes it impractical; the best trade-off in terms of authentication time, robustness and discrimination performance is achieved by the extended GMM approach.

IDIAP2004

Object Duplicate Detection

Péter Vajda

With the technological evolution of digital acquisition and storage technologies, millions of images and video sequences are captured every day and shared in online services. One way of exploring this huge volume of images and videos is through searching a particular object depicted in images or videos by making use of object duplicate detection. Therefore, need of research on object duplicate detection is validated by several image and video retrieval applications, such as tag propagation, augmented reality, surveillance, mobile visual search, and television statistic measurement. Object duplicate detection is detecting visually same or very similar object to a query. Input is not restricted to an image, it can be several images from an object or even it can be a video. This dissertation describes the author's contribution to solve problems on object duplicate detection in computer vision. A novel graph-based approach is introduced for 2D and 3D object duplicate detection in still images. Graph model is used to represent the 3D spatial information of the object based on the local features extracted from training images so that an explicit and complex 3D object modeling is avoided. Therefore, improved performance can be achieved in comparison to existing methods in terms of both robustness and computational complexity. Our method is shown to be robust in detecting the same objects even when images containing the objects are taken from very different viewpoints or distances. Furthermore, we apply our object duplicate detection method to video, where the training images are added iteratively to the video sequence in order to compensate for 3D view variations, illumination changes and partial occlusions. Finally, we show several mobile applications for object duplicate detection, such as object recognition based museum guide, money recognition or flower recognition. General object duplicate detection may fail to detection chess figures, however considering context, like chess board position and height of the chess figure, detection can be more accurate. We show that user interaction further improves image retrieval compared to pure content-based methods through a game, called Epitome.